Ph.D., MHA @PQE Group
Here is what makes this urgent: appropriation does not wait for permission. By the time most organizations formalize an AI strategy, their employees have already appropriated AI on their own terms. Microsoft and LinkedIn’s 2024 Work Trend Index found that 78% of people who use AI at work bring their own tools rather than ones the organization provided [2]. A separate study reported that roughly half of employees use unsanctioned AI tools—and, tellingly, nearly half of those users said they would keep using them even if their employer explicitly banned them [3].
This “bring your own AI” reality is unfaithful appropriation at an organizational scale. Employees aren’t appropriating the spirit of a sanctioned system; they’re improvising with whatever tool(s) relieves their workload, outside any governance the organization can see. And the instinct to suppress it backfires. A ban does not end appropriation—it drives appropriation underground, where it is invisible, ungoverned, and impossible to learn from.
The same research surfaces why this concealment happens. Microsoft and LinkedIn found that a majority of AI users hesitate to admit using AI on their most important tasks, and a comparable share fear that disclosing AI use would make them look replaceable [2]. That fear is not a quirk. It is a signal about perceived organizational support. Organizational Support Theory, established in Part 1, holds that employees calibrate their behavior to whether they believe the organization values them [4]. When the prevailing message is that AI exists to do more with fewer people, the rational response is to hide one’s appropriation—and concealed appropriation is unfaithful appropriation by default.
If appropriation is inevitable and suppression is counterproductive, the organizational task is to steer it toward faithful use. Part 3 introduced the deployment-phase imperative of building feedback infrastructure. The deeper move is to manage the conditions under which appropriation occurs.
Three interventions do this work. First, make the spirit of the tool explicit. Employees appropriate technology faithfully only when they understand what it is for—its intended purpose, its boundaries, and the judgement it is meant to support rather than replace. Where that intent is left implicit, employees infer it, and their inferences drive divergent appropriation. Second, build consensus on appropriation. AST research emphasizes that groups who openly agree on how a tool should be used appropriate it more faithfully and more stably than those who never have the conversation. Surfacing and resolving disagreement about appropriate use is, itself, an intervention. Third, treat unfaithful appropriation as intelligence, not insubordination. When an employee routes around a sanctioned feature, the workaround exposes where the tool does not fit the workflow — data that is only available if the organization has made it safe to disclose.
Steering appropriation at launch is necessary but not sufficient, because appropriation does not stay fixed. Over time, repeated appropriation moves harden into structure: informal practices become “how we do things here,” and those settled patterns are far harder to redirect than fresh ones. This is where many organizations declare victory too early and watch their gains quietly decay.
Actor-Network Theory, the fourth lens from Part 1, explains the longer arc. ANT treats AI not as a static instrument but as an active participant in organizational networks—one that reshapes relationships, redistributes decision-making, and redraws the boundaries of expertise as it becomes embedded in daily work [5]. Once that embedding occurs, the human-AI division of labor keeps shifting, and faithful appropriation established at launch can drift back toward unmonitored workaround if nothing reinforces it.
Preventing that drift is structural, not motivational. Three conditions matter most. Governance must legitimize sanctioned use—providing tools good enough, and policies clear enough, that faithful appropriation is easier than the shadow alternative; where approved options lag, employees revert to BYOAI. Roles and recognition must reflect the new division of labor, so that employees who use AI well are visibly valued rather than quietly made to feel expendable. And stakeholder mapping must be continuous, embedded in operational rhythm rather than treated as a one-time post-mortem. As the work on scaling AI has consistently shown, the organizations that sustain adoption are those that treat it as an ongoing cultural and structural commitment, not a project that ends at go-live [6].
Appropriation is the hinge on which the entire series turns. The organizational support built before launch, the work co-designed during it, the feedback gathered at deployment—all of it pays off, or unravels, in how employees actually appropriate AI once it is in their hands and in whether the organization sustains the conditions for faithful use over time. Adoption can be mandated. Faithful appropriation can only be cultivated.
Which raises the question Part 5 will answer. If appropriation is steered most powerfully by how well a system fits the real contours of work, then the surest way to earn faithful appropriation is to design for it from the start. The final part of this series examines the framework for joint optimization of AI and human work—how to design the technical and human systems together, so that the tool employees receive is one they are inclined to use as intended.
The following steps operationalize the principles above. They are intended for planning and future application:
This article is part of our series on driving successful AI adoption through organizational change. Stay tuned for more insights, and explore the previous pieces below:
References
[1] DeSanctis, G., & Poole, M. S. (1994). Capturing the complexity in advanced technology use: Adaptive structuration theory. Organization Science, 5(2), 121–147. https://doi.org/10.1287/orsc.5.2.121
[2] Microsoft & LinkedIn. (2024). AI at work is here. Now comes the hard part: 2024 Work Trend Index Annual Report. https://www.microsoft.com/en-us/worklab/work-trend-index/ai-at-work-is-here-now-comes-the-hard-part
[3] Software AG. (2024). Shadow AI study (as reported by SecurityWeek, 2025). https://www.securityweek.com/the-shadow-ai-surge-study-finds-50-of-workers-use-unapproved-ai-tools/
[4] Eisenberger, R., & Stinglhamber, F. (2011). Perceived organizational support: Fostering enthusiastic and productive employees. American Psychological Association. https://doi.org/10.1037/12318-000
[5] Walsham, G. (1997). Actor-network theory and IS research: Current status and future prospects. In A. S. Lee, J. Liebenau, & J. I. DeGross (Eds.), Information systems and qualitative research (pp. 466–480). Springer. https://doi.org/10.1007/978-0-387-35309-8_23
[6] Fountaine, T., McCarthy, B., & Saleh, T. (2019). Building the AI-powered organization. Harvard Business Review. https://hbr.org/2019/07/building-the-ai-powered-organization
